Value priority based optimal power system stabilization of generating resources using local and Global Controllers

In this paper, an intelligent supervisory level power system stabilizer based on value based prioritization of Reinforcement Learning (RL) and Supervised Learning (SL) is proposed. The proposed method uses a composite architecture integrating two controllers: First, an Adaptive Critic Design (ACD) implemented on neural network benchmark capable of approximating the nonlinear functional dynamics of the power system and second, a conventional Power System Stabilizer (PSS) as a local controller. The value priority is defined using a Lyuponov function candidate derived based on system stability analysis to indicate identification and performance quality of the controllers. This method increases the reliability and allows for automatic tuning of stabilizing controllers. The theoretical results are validated by conducting simulation studies for electric-generator stabilization on 39-bus 10-generator IEEE power system.